Thanks for the response Aaron.

On Mon, Feb 23, 2009 at 11:34 AM, Aaron Turner <[email protected]> wrote:
> In my experience, sending packets on eth0 causes the packet to bypass
> the TCP/IP stack and be sent out sight unseen.  Hence, you won't be
> able to inject packets into a TCP stream with the target of the local
> host.

Well that kind of sucks. =( It seems weird that an outgoing packet
like this could be filtered by the kernel. What criteria does a packet
like this fit for it to be filtered out? (e.g. it doesn't match a TCP
connection, but obviously it does in this case.) The only thing I can
think of is that the kernel uses a different incoming adapter (really
low level?) than libpcap, which sounds wrong since libpcap should be
operating at the lowest possible layer, right?

> Sending packets via loopback might work- I've never tried that to be
> honest.  I'm not really sure if you can inject standard ethernet
> frames or you need to convert to Linux's cooked SLL header format.
> You might try setting the destination MAC to that of eth0 and see if
> the kernel will route it for you.

Tried sending an incoming packet to lo, no go either. It's the same
packet as if I had sent it on eth0. Should I perhaps be changing the
MAC address to the one on lo? What do packets even look like on the
lo? In that SLL format? I can't think of any application that uses lo
for communication that I can use Windump to look at.

Thanks,
Oliver
-
This is the tcpdump-workers list.
Visit https://cod.sandelman.ca/ to unsubscribe.

Reply via email to